You should be able to autosuggest bits of the program based on what has been collected so far. (All IDEs with "suggestions" basically do this).
Using the syntax of the assembler, anyplace a partial code fragment exists, a "syntax error corrector" can propose bits to insert or delete to fix the syntax. Variations of this include completing an instruction mnemonic, especially based on frequency of static occurrence of instructions in code, or typical code orders (just coded a CMP? Pretty sure a jump conditional is next). You might also suggest or autofill the rest of an addressing mode; many start with an opening paren or bracket, and after an operand, must be closed with the corresponding paren.
You can also collect the set of identifiers in the program, and using spelling correcting techniques (suffix trees, levehstein distance), propose the rest of an identifier. (You have no idea how often I've wished for this in a 30,000 line assembler application I continue to maintain).
Now, these are all (natch!) great ideas. Whether they are in the scope of things you think you know how to do, or are willing to do, is another question entirely. What this is likely to take is a considerable set of machinery for parsing, collecting/correcting symbols, analyzing partial programs, etc.; this is pretty daunting to build. (I've been trying to do something like this for the last 20 years with modest success, see my bio for more details).